In vivo near-infrared fluorescence imaging

https://doi.org/10.1016/j.cbpa.2003.08.007Get rights and content

Abstract

Photon penetration into living tissue is highly dependent on the absorption and scattering properties of tissue components. The near-infrared region of the spectrum offers certain advantages for photon penetration, and both organic and inorganic fluorescence contrast agents are now available for chemical conjugation to targeting molecules. This review focuses on those parameters that affect image signal and background during in vivo imaging with near-infrared light and exogenous contrast agents. Recent examples of in vivo near-infrared fluorescence imaging of animals and humans are presented, including imaging of normal and diseased vasculature, tissue perfusion, protease activity, hydroxyapatite and cancer.

Introduction

This review focuses on recently published examples of in vivo near-infrared (NIR) fluorescence imaging. To be successful, each example addressed a common set of parameters (outlined below and summarized in Box 1) and optimized the signal to background ratio (SBR; also called signal to noise ratio).

The use of NIR (700–1000 nm) light for biomedical imaging is grounded in first principles, and is best understood in the context of photon propagation through living tissue and the SBR. An excitation photon typically travels through tissue to reach the fluorescent contrast agent, and has several possible fates depending on the tissue’s scatter, anisotropy (g), and refractive index(ices) [1]. The photon emitted by the fluorophore is subject to the same fates. Generally, the photon absorbance of a particular tissue or organ is the sum of all absorbing components present. In living, non-pigmented tissue, the major NIR absorbers are water, lipids, oxyhemoglobin and deoxyhemoglobin, with the absolute value of μA depending on the molar concentration of each component 2., 3.••. In a ‘typical’ tissue, having an 8% blood volume and 29% lipid content, the dominant absorber is hemoglobin, accounting for 39–64% of total absorbance at NIR wavelengths [3••].

Whenever tissue absorbs light, there is a chance that fluorescent light will be emitted. In addition to absorption, tissue ‘autofluorescence’ can severely limit SBR. Figure 1 demonstrates the relationship of excitation and emission wavelengths to tissue autofluorescence. ‘Green’ autofluorescence of the skin and viscera, and especially the gallbladder, small intestine and bladder, is astoundingly high when excited with blue light. Even a small amount of urine leaking from the urethra (overlying the tail) is easily seen. Autofluorescence of the gallbladder and bladder are markedly reduced using a ‘red’ filter set (green light excitation), but intestinal autofluorescence remains significant. Use of a NIR filter set essentially eliminates autofluorescence. Hence, high tissue autofluorescence precludes the use of visible light for most in vivo imaging applications, and NIR light solves this problem by reducing fluorescence background.

For reflectance-type imaging (see below), scatter is as important as absorbance with respect to photon attenuation. Scatter describes the deviation of a photon from the parallel axis of its path, and can occur when the tissue inhomogeneity is small relative to wavelength (Rayleigh scattering), or roughly equal to wavelength (Mie scattering). Empirical measurements suggest that the wavelength dependence of the scatter coefficient is strongly dependent on tissue composition. In rat skin, the reduced scattering coefficient is proportional to λ−2.4 [4], suggesting Rayleigh dominance; however, in post-menopausal human breast it is proportional to λ−0.6 [5], suggesting Mie dominance. A detailed analysis of how tissue absorption and scatter affect the selection of excitation and emission wavelengths for reflectance in vivo NIR fluorescence imaging has recently been published [3••].

Because there is little NIR fluorescence contrast generated by most tissues, most in vivo studies administer exogenous contrast agents. Until recently, exogenous contrast was limited to organic fluorophores, the most common of which are polymethines. One important class of these molecules is the heptamethine cyanines (Figure 2a), comprising benzoxazole, benzothiazole, indolyl, 2-quinoline and 4-quinoline subclasses. Of these, the indocyanines are the most widely used as they do not have the aggregation problems associated with the other subclasses [6]. Peak excitation of this class is at 760–800 nm, and peak emission at 790–830 nm.

In 1958, indocyanine green (ICG; Cardio-Green; Figure 2b) was submitted for approval to the FDA for use in indicator-dilution studies in humans. It is one of the least toxic agents ever administered to humans, with the only known adverse reaction being rare anaphylaxis. Recently, several improved heptamethine indocyanines have become available. For example, Figure 2c shows the carboxylic acid form of IRDye78 (LI-COR, Lincoln, NE). This molecule is tetra- rather than di-sulfonated, which increases aqueous solubility and aqueous quantum yield (QY), and is available as an N-hydroxysuccinimide ester derivative for covalent conjugation to targeting molecules. Dozens of other heptamethine fluorophores have been reported 7., 8.•, 9.. Conjugation to the indocyanines requires careful handling [10], and a set of robust methods has recently been published [11].

Conventional organic fluorophores suffer from significant limitations. First, and foremost, it is difficult to control excitation and emission wavelengths. These wavelengths are dependent on chemical structure, and tuning a conventional fluorophore to precise wavelengths requires sophisticated chemistry. Moreover, changes in wavelength will usually be ‘discrete’. For polymethines, the addition of each double bond results in an emission wavelength increase of 80–100 nm, but comes at the cost of a more hydrophobic and potentially unstable molecule. Second, QY is usually less than 15% in aqueous environments. Third, for conjugation of tumor targeting, or other ligands to fluorophores, there is usually one ligand per fluorophore, so detectability is a strong function of target concentration. Finally, conventional fluorophores are highly susceptible to photobleaching, which limits the fluence rate that can be applied to a sample, and hence the sensitivity of detection (see below).

Inorganic fluorescent semiconductor nanocrystals (also called quantum dots; QDs) have the potential to solve many of these problems. QDs are synthesized in organic solvents such as hexane, and typically comprise an inorganic core and inorganic shell of metal, the properties of which tune peak fluorescence emission. For in vivo imaging, QDs must be additionally coated with an aqueous-compatible organic layer (Figure 2d). QD fluorescence emission can be tuned to virtually any discrete wavelength and absorption is broadband. More importantly, QDs are remarkably resistant to photobleaching, permit dozens of targeting molecules to be conjugated to a single QD, and, at least in organic solvents, have very high QYs (discussed below). A detailed discussion of QD properties for in vivo imaging has been published [3••].

Solubility is a significant issue for both organic and inorganic fluorophores. Non-sulfonated heptamethines, such as IR-786, have aqueous solubilities ≤ 10 μM, and require carriers such as Cremophor EL and/or ethanol for in vivo use [12]. Solubility increases dramatically for every charged group, typically sulfonates, substituted on the fluorophore. For the heptamethine indocyanines, tetra-sulfonation typically results in aqueous solubilities of ≥ 10 mM. Of course, the cost of adding too much charge is that some targeting molecules will lose binding potency because of charge or steric effects. As a general rule, the degree of sulfonation should be chosen to render the NIR fluorophore/targeting molecule conjugate soluble in plasma, but not so charged as to interfere with target binding. For QDs, solubility depends entirely on the organic coating applied to the inorganic core/shell. Although this field is in its infancy, several organic coatings have been described 3.••, 13., 14., 15., 16., 17..

Heptamethine indocyanines typically have extinction coefficients in aqueous buffers of 100 000–200 000 M−1cm−1, which, when coupled with low tissue absorption in the NIR, results in reasonably good photon absorption by the fluorophore. QDs possess the interesting property of broadband absorption; they absorb all wavelengths below their emission wavelength, and their absorption coefficient actually increases as excitation wavelength becomes bluer. This property permits multiple QDs with different fluorescence emissions to be excited by a single wavelength, or even broadband light, although the absorption and scattering properties of tissue can negate much of this advantage [3••].

Absolute QY can be difficult to measure, and most reports of QY are given for methanol or other non-aqueous solvents. In general, the QY of heptamethine indocyanines in aqueous environments increases with their solubility, but even for tetra-sulfonated molecules is often less than 15% [12]. A recent report of novel heptamethine indocyanines with a 28% aqueous QY suggests that improvements are possible [9]. QD QYs, too, are typically cited for organic solvents, where QY can approach 90%. However, some of the newer aqueous-solubilizing organic coatings for QDs have been reported to produce QYs of 50–60%.

A major difference favoring QDs over organic fluorophores is their resistance to photobleaching. In our experience, heptamethine indocyanines rapidly photobleach in serum when fluence rate exceeds 50 mW/cm2 [12]. This effect precludes the use of higher fluence rates, which would otherwise improve the SBR. QDs can withstand fluence rates one to two orders of magnitude higher. Because 50 mW/cm2 is already a sizable fluence rate, and tissue damage can occur at high fluence rates, it is unclear how much of an issue this will become for in vivo imaging.

In vivo imaging requires that the contrast agent is delivered to the target, has adequate contact time with the target for binding to occur, and is retained by the target while non-bound material is cleared from the circulation. Unconjugated organic NIR fluorophores are typically ≤ 1200 Da, but have widely different biodistributions and pharmacokinetics depending on their charge and the properties of their conjugated targeting molecule. In the absence of any solubilizing groups, heptamethine indocyanines are lipophilic cations that distribute freely and partition intracellularly. This property can been exploited to create a perfusion tracer 12., 18.•. As charge (e.g. sulfonation) is increased, fluorophores remain extracellular and plasma half-life increases proportionally [8]. Even so, the plasma half-life of unconjugated tetra-sulfonated heptamethine indocyanines is measured in minutes at best. Clearance is typically accomplished from a combination of renal filtration and excretion into bile.

QDs have the opposite problem. Because of their multi-layered structure, and especially given the thickness of the organic coatings, they are typically large structures of 3–20 nm hydrodynamic diameter (HD). Because there is a sharp decrease in renal filtration above a 3.5 nm HD, most QDs will be difficult to clear from the circulation, resulting in high background. This problem has yet to be addressed and solved systematically.

Without charged groups, heptamethine indocyanines can be quite toxic due to intracellular accumulation [12]. However, di-sulfonated ICG has been used routinely in humans for over 40 years and has an excellent safety profile. Less is known about other heptamethine indocyanines, although studies in mice suggest a dramatic increase in the LD50 with increasing sulfonation [8].

The in vivo toxicity, if any, of QDs is unknown. They typically comprise metals that, in their elemental form, are toxic (e.g. cadmium, arsenic and lead). However, in the nanocrystalline form, these metals are already bound as a less reactive salt, and might be relatively inert in the body. This critical issue remains to be addressed.

The Weissleder group has pioneered the concept of ‘stealth’ NIR fluorescent probes [19]. These contrast agents are ‘silent’ in their native form by virtue of a quenching molecule connected via a short linker. After proteolysis or chemical scission of the linker, the contrast agent becomes unquenched and highly fluorescent. This system results in extremely low background because the majority of the injected dose is non-fluorescent. Because the linker is typically hydrolyzed by an enzyme, the system also produces high signal via ‘amplification’. Examples of in vivo imaging with this system are given below. Amplification can also occur via endocytosis of a contrast agent bound to the cell surface, although the stability, over time, of NIR fluorophores in the endosomal compartment is unknown. It may also be possible to apply stealth technology to QDs given their relatively long fluorescent lifetime.

There are two fundamental configurations for in vivo NIR fluorescence imaging: reflectance and tomographic. In reflectance imaging, continuous wave excitation light is delivered to the surface of the subject, photons traverse the boundary layer and reach the target, and emitted light travels back to the surface where it is collected directly. Although imaging is instant, there is exponential decay of intensity with tissue depth, and scatter limits the technique to several millimeters, at best. In tomographic imaging, signal intensity falls more linearly with target depth in tissue, permitting measurements to be made up to 30–40 cm into tissue 20., 21.. However, the image must be reconstructed from multiple measurements made in the time or frequency domains using variants of the diffusion equation. Two excellent recent reviews describe the nuances of reflectance versus tomographic imaging 22.••, 23.••.

Regardless of the imaging system employed, one must be extremely careful in selecting filters, lenses, cameras and other optics for in vivo NIR fluorescence imaging. Most optics are optimized only for visible light transmission, and many have a heat coating that will drastically reduce NIR transmission. For imaging, silicon CCD cameras are preferred because of their cost/performance ratio; however, sensitivity is fair to poor above 800 nm, and few silicon cameras have quantum efficiencies higher than 25% at 800 nm. The quantum efficiency of gallium arsenide is excellent up to 900 nm; however, cost is higher and availability limited. Indium gallium arsenide cameras are not useful for NIR imaging because of their poor quantum efficiency below 1000 nm, and higher noise.

Section snippets

Recent advances in in vivo near-infrared fluorescence imaging

Over the past several years, there has been an explosion of reports describing successful in vivo NIR fluorescence imaging. Many studies are qualitative; however, quantitative methods are beginning to emerge where dictated by necessity. The following summary describes those studies in the context of the principles discussed above.

Conclusions

Organic and inorganic fluorophores that span the entire NIR portion of the spectrum are now available. When conjugated covalently to targeting molecules, sensitive contrast agents for NIR fluorescence imaging are created. However, to be used successfully in vivo, careful attention must be paid to those parameters that influence the SBR, including selection of excitation and emission wavelengths, fluorophore solubility, QY, photobleaching threshold, biodistribution, target binding affinity,

References and recommended reading

Papers of particular interest, published within the annual period of review, have been highlighted as:

  • of special interest

  • ••

    of outstanding interest

Acknowledgements

I thank Ralph Weissleder (Massachusetts General Hospital) and Moungi G Bawendi (Massachusetts Institute of Technology) for critical reading of this manuscript, Alec M DeGrand for technical assistance, and Grisel Rivera for administrative assistance. This work was supported by Department of Energy (Office of Biological and Environmental Research) grant #DE-FG02-01ER63188, a Clinical Scientist Development Award from the Doris Duke Charitable Foundation, and grants from the National Institutes of

References (55)

  • Y.T. Lim et al.

    Selection of quantum dot wavelengths for biomedical assays and imaging

    Mol. Imaging

    (2003)
  • W.F. Cheong et al.

    A review of the optical properties of biological tissues

    IEEE J. Quantum Electron

    (1990)
  • A.E. Cerussi et al.

    Sources of absorption and scattering contrast for near-infrared optical mammography

    Acad. Radiol.

    (2001)
  • J.H. Flanagan et al.

    Functionalized tricarbocyanine dyes as near-infrared fluorescent probes for biomolecules

    Bioconjug. Chem.

    (1997)
  • Miwa N, Inagaki M, Eguchi H, Okumura M, Inagaki Y, Harada T: Near-infrared fluorescent contrast agent and fluorescence...
  • Y. Lin et al.

    Novel near-infrared cyanine fluorochromes: synthesis, properties, and bioconjugation

    Bioconjug. Chem.

    (2002)
  • R. Rajagopalan et al.

    Stabilization of the optical tracer agent indocyanine green using noncovalent interactions

    Photochem. Photobiol.

    (2000)
  • A. Zaheer et al.

    IRDye78 conjugates for near-infrared fluorescence imaging

    Mol. Imaging

    (2002)
  • A. Nakayama et al.

    Quantitation of brown adipose tissue perfusion in transgenic mice using near-infrared fluorescence imaging

    Mol. Imaging

    (2003)
  • M. Bruchez et al.

    Semiconductor nanocrystals as fluorescent biological labels

    Science

    (1998)
  • W.C.W. Chan et al.

    Quantum dot bioconjugates for ultrasensitive nonisotopic detection

    Science

    (1998)
  • H. Mattoussi et al.

    Self-assembly of CdSe-ZnS quantum dot bioconjugates using an engineered recombinant protein

    J. Am. Chem. Soc.

    (2000)
  • X. Wu et al.

    Immunofluorescent labeling of cancer marker Her2 and other cellular targets with semiconductor quantum dots

    Nat. Biotechnol.

    (2003)
  • B. Dubertret et al.

    In vivo imaging of quantum dots encapsulated in phospholipid micelles

    Science

    (2002)
  • A. Nakayama et al.

    Functional near-infrared fluorescence imaging for cardiac surgery and targeted gene therapy

    Mol. Imaging

    (2002)
  • R. Weissleder et al.

    In vivo imaging of tumors with protease-activated near-infrared fluorescent probes

    Nat. Biotechnol.

    (1999)
  • V. Ntziachristos et al.

    Would near-infrared fluorescence signals propagate through large human organs for clinical studies?

    Opt. Lett.

    (2002)
  • Cited by (2302)

    View all citing articles on Scopus
    View full text